Figure 3

Flowchart depicting Oculus behavior with example sequences. As input is parsed, new sequences are passed into the aligner in the order they are observed. The aligner then performs normally, mapping each passed read to the database. Downstream of the aligner, Oculus expands the alignment file to reflect the count of each input sequence. Since compression and reconstitution are faster than alignment, there is a net reduction in runtime. In reverse-complement mode (Section 2.4), Oculus would remove the read sequence TTTT, having already seen AAAA, and print an additional alignment: chr-001 with reversed orientation. By default, Oculus treats AAAA and TTTT as distinct sequences – both would be passed into the aligner.